<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
wav2vec2-large-xlsr-mecita-coraa-portuguese-all-04
This model is a fine-tuned version of Edresson/wav2vec2-large-xlsr-coraa-portuguese on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.1868
- Wer: 0.0906
- Cer: 0.0281
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 3e-05
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 100
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 28.5317 | 1.0 | 86 | 3.4264 | 1.0 | 1.0 |
| 8.6384 | 2.0 | 172 | 3.0049 | 1.0 | 1.0 |
| 3.0821 | 3.0 | 258 | 2.9416 | 1.0 | 1.0 |
| 2.9357 | 4.0 | 344 | 2.9331 | 1.0 | 1.0 |
| 2.9035 | 5.0 | 430 | 2.9165 | 1.0 | 1.0 |
| 2.8816 | 6.0 | 516 | 2.6902 | 1.0 | 1.0 |
| 2.3317 | 7.0 | 602 | 1.0713 | 0.7650 | 0.1968 |
| 2.3317 | 8.0 | 688 | 0.5039 | 0.2577 | 0.0686 |
| 1.0095 | 9.0 | 774 | 0.3925 | 0.2034 | 0.0550 |
| 0.6723 | 10.0 | 860 | 0.3236 | 0.1740 | 0.0477 |
| 0.5441 | 11.0 | 946 | 0.3129 | 0.1486 | 0.0430 |
| 0.4353 | 12.0 | 1032 | 0.2682 | 0.1429 | 0.0414 |
| 0.391 | 13.0 | 1118 | 0.2518 | 0.1316 | 0.0384 |
| 0.3714 | 14.0 | 1204 | 0.2512 | 0.1244 | 0.0363 |
| 0.3714 | 15.0 | 1290 | 0.2382 | 0.1197 | 0.0356 |
| 0.3316 | 16.0 | 1376 | 0.2245 | 0.1155 | 0.0351 |
| 0.3111 | 17.0 | 1462 | 0.2191 | 0.1089 | 0.0332 |
| 0.2925 | 18.0 | 1548 | 0.2103 | 0.1044 | 0.0317 |
| 0.2865 | 19.0 | 1634 | 0.2053 | 0.1061 | 0.0324 |
| 0.2679 | 20.0 | 1720 | 0.2100 | 0.1002 | 0.0323 |
| 0.2429 | 21.0 | 1806 | 0.1974 | 0.0995 | 0.0312 |
| 0.2429 | 22.0 | 1892 | 0.2043 | 0.0975 | 0.0307 |
| 0.2421 | 23.0 | 1978 | 0.2041 | 0.0973 | 0.0306 |
| 0.2366 | 24.0 | 2064 | 0.2022 | 0.0965 | 0.0300 |
| 0.2262 | 25.0 | 2150 | 0.1995 | 0.0990 | 0.0306 |
| 0.2084 | 26.0 | 2236 | 0.1954 | 0.0992 | 0.0306 |
| 0.2204 | 27.0 | 2322 | 0.1966 | 0.0987 | 0.0298 |
| 0.2151 | 28.0 | 2408 | 0.1983 | 0.0965 | 0.0297 |
| 0.2151 | 29.0 | 2494 | 0.1970 | 0.0913 | 0.0285 |
| 0.1998 | 30.0 | 2580 | 0.1980 | 0.0938 | 0.0285 |
| 0.1944 | 31.0 | 2666 | 0.1871 | 0.0943 | 0.0289 |
| 0.1929 | 32.0 | 2752 | 0.2006 | 0.0926 | 0.0281 |
| 0.2138 | 33.0 | 2838 | 0.1925 | 0.0913 | 0.0286 |
| 0.1913 | 34.0 | 2924 | 0.1921 | 0.0916 | 0.0285 |
| 0.1842 | 35.0 | 3010 | 0.1949 | 0.0913 | 0.0284 |
| 0.1842 | 36.0 | 3096 | 0.1938 | 0.0938 | 0.0294 |
| 0.1906 | 37.0 | 3182 | 0.1944 | 0.0938 | 0.0295 |
| 0.182 | 38.0 | 3268 | 0.1920 | 0.0889 | 0.0284 |
| 0.1729 | 39.0 | 3354 | 0.1932 | 0.0911 | 0.0290 |
| 0.1766 | 40.0 | 3440 | 0.1868 | 0.0906 | 0.0281 |
| 0.1789 | 41.0 | 3526 | 0.1914 | 0.0903 | 0.0282 |
| 0.1595 | 42.0 | 3612 | 0.1871 | 0.0894 | 0.0280 |
| 0.1595 | 43.0 | 3698 | 0.1954 | 0.0886 | 0.0274 |
| 0.1632 | 44.0 | 3784 | 0.1868 | 0.0901 | 0.0280 |
| 0.1686 | 45.0 | 3870 | 0.2000 | 0.0866 | 0.0271 |
| 0.1577 | 46.0 | 3956 | 0.1891 | 0.0884 | 0.0276 |
| 0.1572 | 47.0 | 4042 | 0.1941 | 0.0886 | 0.0277 |
| 0.1581 | 48.0 | 4128 | 0.1949 | 0.0876 | 0.0270 |
| 0.154 | 49.0 | 4214 | 0.1942 | 0.0899 | 0.0279 |
| 0.1519 | 50.0 | 4300 | 0.1938 | 0.0886 | 0.0272 |
| 0.1519 | 51.0 | 4386 | 0.1979 | 0.0908 | 0.0286 |
| 0.1494 | 52.0 | 4472 | 0.2006 | 0.0889 | 0.0277 |
| 0.1516 | 53.0 | 4558 | 0.2010 | 0.0906 | 0.0287 |
| 0.1353 | 54.0 | 4644 | 0.2045 | 0.0901 | 0.0286 |
| 0.1473 | 55.0 | 4730 | 0.1949 | 0.0899 | 0.0282 |
| 0.1417 | 56.0 | 4816 | 0.1923 | 0.0906 | 0.0289 |
| 0.1398 | 57.0 | 4902 | 0.1923 | 0.0906 | 0.0283 |
| 0.1398 | 58.0 | 4988 | 0.1960 | 0.0894 | 0.0279 |
| 0.1449 | 59.0 | 5074 | 0.1964 | 0.0913 | 0.0283 |
| 0.1412 | 60.0 | 5160 | 0.1998 | 0.0899 | 0.0280 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.1+cu117
- Datasets 2.12.0
- Tokenizers 0.13.3
